Original Date: 01/23/1995
Revision Date: 01/18/2007

Sandia is developing low volume statistical process control (LVSPC) to apply to its small quantity of products produced. The method used to evaluate the process involves an algorithm which updates the estimate of a time-varying mean whenever more data becomes available. This type of estimate is more effective than the usual sample average when the process mean varies over time.

The model used for low-volume production is an extension of the classical Shewhart approach which models the process mean as varying over time, and leads to an estimate of the current mean weighted to the most recent data. Two types of control charts with increased sensitivity for LVSPC include the cumulative sum and exponential weighted moving average methods. These charts use information from the entire sequence of available data points with points nearest to the current point weighted to have the greatest effect on the process mean. Additional work with these models has been used to determine the number of points necessary to achieve statistical validity. This number is based on the process parameters and the amount of noise and variation of noise in a particular process.

Sandia is working to establish an applicable SPC tool when data is limited. Theoretical statistical research with some limited production for validation has aided in developing methods for LVSPC. Integration of new and traditional SPC methods with adaptive filters is providing new and improving tools for low volume process controls. LVSPC will prove beneficial to industry as industry becomes more agile and moves towards smaller product quantities. Useful and accurate LVSPC techniques are needed because of the scarcity of process data in low volume production.

For more information see the Point of Contact for this survey.